testing-llm
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides comprehensive guidelines for testing AI features using well-known and trusted libraries such as DeepEval, RAGAS, and VCR.py.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill proactively recommends security measures to prevent data exposure. Specifically, it instructs users to filter sensitive headers (e.g., 'authorization', 'x-api-key') when recording network traffic and explicitly warns against hardcoding API keys or committing them to version control.
- [EXTERNAL_DOWNLOADS]: The skill references and suggests the installation of standard, well-known Python packages (deepeval, ragas, pytest, vcrpy, pydantic, scipy) and Node.js tools (playwright). All external documentation links point to official and well-known service domains (deepeval.com, ragas.io, playwright.dev, scipy.org).
- [INDIRECT_PROMPT_INJECTION]: While the skill discusses testing for prompt injection in applications, it does not introduce such vulnerabilities itself. Instead, it promotes structured output validation using Pydantic, which is a best practice for sanitizing and validating untrusted LLM outputs.
- [DYNAMIC_EXECUTION]: The described 'Healer Agent' workflow for fixing end-to-end tests is a standard feature of the Playwright framework and does not involve unsafe dynamic code generation or execution within the skill's own logic.
Audit Metadata